How to run Unified Detection¶

What can I do with Unified Detection?

Create consistent object proposals across a dataset using both VLM and SAM.
Merge overlapping detections from multiple sources.
Prepare inputs for downstream operators such as Find Similar or Modify Detections.

1. Select your dataset¶

To load a dataset for the first time, use the Create Dataset from Directory operator.

Create Dataset from Directory

Give the dataset a name, provide a relative or absolute path to the dataset directory, and click Execute.
Once you've loaded a dataset, it will always be available to load in the Select Dataset menu.

Load Dataset

Quick Load Dataset

2. Launch the `Unified Detection` operator¶

Select the Unified Detection operator in the Browse Operations menu.

Select Unified Detection

3. Choose the detection mode¶

Find Description – detect only prompted classes using a Vision Language Model (VLM).
Find Everything – detect all visible objects using Segment Anything Model (SAM).
Both (Find Description + Find Everything) – run both modes and merge results (recommended for general use).

4. Enter prompts (for VLM)¶

Supply comma-separated object names or short phrases to guide Find Description:

pallet jack, forklift, hard hat

If needed, select your GPU device, then click Execute.

Execute Unified Detection

5. Adjust parameters (if needed)¶

When using Find Everything, Max Object Area and Min Object Area often need to be adjusted to get the right fit for your dataset.

Max Object Area is a value between 0.0 and 1.0 that defines the maximum allowed size of detected objects as fraction of image area (e.g., 0.2 = 20%).
Min Object Area is a value between 0.0 and 1.0 that defines the minimum allowed size of detected objects as fraction of image area (e.g., 0.003 = 0.3%).

Once you've made your adjustments, click Execute to rerun Unified Detection.

Adjust Parameters

Adjust Unified Detection parameters

Info

For a complete list of operator parameters, consult the reference page.

6. Inspect results¶

Once Unified Detection has finished running, open a sample and confirm the presence of vlm_detections, sam_bbox, sam_mask, and unified_detections (as applicable), as well as leip_embedding.

Inspect detections

Inspect Unified Detections

Troubleshooting¶

If you run into problems, consult the troubleshooting guide.